Cerebionics builds an EEG command channel for robots and drones

Why it matters

A wearable EEG system could give people operating robots, drones or other machines another way to issue commands. Cerebionics will need to show that its decoder works consistently across users and can do so with minimal calibration.

Cerebionics founder and CEO Agnessa Pedersen is building a non-invasive brain-computer interface (BCI) that turns EEG signals into commands for machines. In a September 24th profile, Tech.eu reported that the system had spent months being tested and refined with users in Ukraine. Pedersen's idea grew from a teenage fascination with robots that could draw; now she is working to connect human intent to machines.

Pedersen described a platform that uses EEG electrodes to detect broad patterns associated with intended actions. It targets higher-level commands such as changing direction, confirming an action or aborting it. Cerebionics says its system can connect to drones, unmanned ground vehicles and command-and-control software.

Scalp-based EEG avoids surgery and implanted electrodes, while capturing a lower-resolution signal. Cerebionics is betting that many machine interfaces can work with a limited set of commands rather than a detailed reconstruction of movement. The system has to recognize what a user intends to do and pass that instruction along reliably.

From drawing robots to intent

Pedersen told Tech.eu she began working on brain-computer interfaces in 2024, while studying for an integrated master's in engineering focused on cybernetics and robotics. She had first built robotic arms after taking online courses in Python and C++. Her original plan was to let the robots sketch the first layers of large drawings while she was at school, then finish the work herself.

That hobby led her toward the interface between people and machines. Pedersen experimented with flex-sensor gloves, but found they measured movement as it happened. She wanted to detect intent earlier, before physical execution. "I wanted to connect our minds to machines and remove the need for physical control," she told Tech.eu.

After Project Europe became involved roughly a year before the interview, Pedersen moved to London and began working on Cerebionics full-time, she said. Dealroom's company record lists a €200,000 seed investment from Project Europe in July 2025. That is an earlier financing milestone, not a newly announced round.

The hard part is making it work across people

Cerebionics describes potential uses across defence, healthcare, industry and consumer applications. Pedersen said the system can recognize high-level commands and that the team is expanding the set it can interpret. She also described a safety layer intended to prevent action when the model's confidence is low.

Pedersen identified software as the harder development problem. The hardware can capture workable signals, she told Tech.eu; the challenge is the decoder and making the platform generalizable and transferable. Cerebionics currently requires calibration because a person's signal patterns can change from day to day. For broader adoption, Pedersen said the platform will need to work with minimal calibration.

Response time depends on more than the decoder. Signal capture, classification, communications and the receiving machine all contribute. As Pedersen noted, a brain command to a drone still depends on the connection between the headset and that drone remaining usable, including in environments where communications may be disrupted.

A different wager on brain-computer interfaces

The comparison with Neuralink can obscure the product Cerebionics is trying to build. Neuralink's implanted approach seeks higher-resolution signals for applications involving people with severe conditions. Cerebionics is pursuing a wearable, non-invasive interface aimed at a smaller vocabulary of machine commands. EEG makes the system easier to wear without surgery, while giving the software less detailed signals to interpret.

Cerebionics also enters a field with other non-invasive EEG developers. Cogitat, for example, develops software to decode EEG signals into digital commands. Cerebionics' account emphasizes machine control and the possibility of using a headset alongside voice, eye tracking and muscle sensors.

A BCI could add a way to confirm, switch or abort while a person's hands and eyes are occupied, without replacing a joystick or voice interface. Whether that becomes a commercial product depends on making the decoder transferable, reducing calibration and showing that it can distinguish deliberate commands from background brain activity consistently.

Pedersen's original instinct was to remove friction between a person's idea and a machine's action. Cerebionics has a defined use case; its software and calibration requirements remain central engineering challenges.